DomainKey-Signature: a=rsa-sha1; c=nofws; d=wiggum.com; h=cc:message-id:from:
	to:in-reply-to:content-type:content-transfer-encoding:
	mime-version:subject:date:references; q=dns; s=mail; b=R8nWwDHvV
	m6Ssqiq5OQtwkERrhwmCuptBFrZwGrj8OvLOjd40XcLw4V5MU53iN7YbSKXmMyug
	HOfiLHiQSeOSv6wuQFsaZ7zvL8FO7UcErebLIFEhKdDfV0p9oUTqJycBLb4Kh4Ui
	xiiaY+I1Yz1YMBGypiWJcy4ZmiYkE2t07U=
Cc: "pgsql-performance" <pgsql-performance@postgresql.org>
Message-Id: <DF26F4B4-FCF2-40DC-89B7-859854AC628F@wiggum.com>
From: Sean Leach <sleach@wiggum.com>
To: Stephen Denne <Stephen.Denne@datamail.co.nz>
In-Reply-To: 
 <F0238EBA67824444BC1CB4700960CB4804B0C9B6@dmpeints002.isotach.com>
Content-Type: text/plain; charset=US-ASCII; format=flowed; delsp=yes
Content-Transfer-Encoding: 7bit
Mime-Version: 1.0 (Apple Message framework v919.2)
Subject: Re: Weird issue with planner choosing seq scan
Date: Mon, 25 Feb 2008 15:32:10 -0800
References: <F0238EBA67824444BC1CB4700960CB4804B0C93A@dmpeints002.isotach.com>
	<C2823FCB-7B73-4838-B792-E1B73C65A254@wiggum.com>
	<F0238EBA67824444BC1CB4700960CB4804B0C9B6@dmpeints002.isotach.com>


On Feb 25, 2008, at 2:59 PM, Stephen Denne wrote:
>>
>
> Please know that I'm very new at advising PostgreSQL users how they  
> should tune their system...

I'd never have known it if you hadn't said anything

>
>
> My understanding of your vacuum verbose output was that it was  
> pointing out that max_fsm_pages was currently smaller than 281727,  
> so therefore there was no way it could contain mappings to all the  
> reusable space. However I don't think it is hinting at, nor  
> recommending a value that you should be using.
>
> If you do nothing, then this number of pages with reusable space  
> will probably continue to grow, therefore, it probably has been  
> growing.
>
> So, for example, if your max_fsm_pages is currently only 20000, then  
> perhaps 20000 of the 281727 pages with reusable space are in the  
> free space map. The remaining 260000 pages _may_ have been generated  
> through 20 different processes each of which created 13000 more  
> pages with reusable space than the map could reference. If that was  
> the case, then a max_fsm_pages of 33000 might be large enough.
>
> Do you see what I'm getting at?
> I think that you should do a vacuum full of that table once, then  
> monitor the number of pages in it with reusable space for a while  
> (over a few iterations of your regular processes). That should give  
> you information about how much larger your max_fsm_pages should be  
> than it currently is.

This sounds sane to me, will do.  Thanks again!